Vice President of Vehicle Engineering:

Good afternoon, everyone and welcome to Tesla's Third Quarter 2021 Q&A webcast. My name is Martin Viecha, Senior Director of Investor Relations and I'm joined today by our CFO, Zachary Kirkhorn, and our Senior VP, Andrew Baglino, as well as other executives. Our Q3 results were announced at about 3 p.m. Central Time in the update deck we published at the same link as this webcast. During this call, we will discuss our business outlook and make forward-looking statements. These comments are based on our predictions and expectations as of today. Actual events or results could differ materially due to a number of risks and uncertainties, including those mentioned in our most recent filings with the SEC. During the question-and-answer portion of today's call, please limit yourself to one question and one follow-up. (Operator instructions) But before we jump into the Q&A, Zach has some opening remarks. Zach?

Yeah. Thanks, Martin. We're continuing to make great progress as a company, setting new records on each of the most important financial metrics for Q3. Overall, we delivered just over 240,000 cars, 20% higher than last quarter, and 70% higher than the same quarter last year. We were also able to achieve an annualized production run rate of over 1 million cars towards the end of the quarter. The increase in production rate has primarily been driven by further ramping of the Model Y at our Shanghai factory. Additionally, we have made great progress, increasing production volumes of Model S, and have recently started to ramp deliveries of Model X. It will take a bit more time to get this program back to prior volumes. But based on demand, we're targeting to exceed historical production levels. We've also completed the transition of our Shanghai factory as our main export hub. This has enabled us to supply more vehicles to the North America market and to introduce Model Y to Europe. Due to part shortages and logistics variability, we have not been able to run our factories at full capacity. It's important to note that while we have roughly doubled deliveries year-to-date, this has been exceptionally difficult to achieve. I want to thank our supply chain team for their incredible work and our production teams for showing impressive flexibility as we make adjustments in real time. These teams' expertise in the chip industry across all tiers has made a huge difference in managing through these challenges. Additionally, we never reduced our production forecast with our suppliers as we're adding capacity as quickly as possible. I also want to thank our suppliers for the dedication and partnership to Tesla. Despite these increases in production and generally higher prices, our backlogs are continuing to grow and average customer wait times are extending. The only practical way to address this in the immediate term is to do everything we can to build more cars on our existing production lines, which is where we are focused. Similar dynamics are also playing out in our storage business as we are working to expand Powerwall and Megapack production as quickly as parts themselves allow us to do so. Additionally, we have made good progress on the in-house battery manufacturing program. And we're excited to have expanded the full self-driving beta program to more customers. Financially, our auto gross margins reached 30.5% on a GAAP basis and just under 29%, excluding regulatory credits, which was our strongest yet. This benefit primarily comes from higher volumes, particularly out of the Shanghai factory, increased mix of the Model Y, and we have made good progress increasing Model S volumes. The Model S has now returned to positive gross margin and we expect this to increase with higher production and the ramp of Model X. As was the case in Q2, there were some net benefit from pricing actions. However, this remains small in the context of other contributors. Please keep in mind that given the backlog, it will take time for the impact of recent changes to flow through our financials. Note that we are also not yet recognizing additional revenue from the FSD Beta program. Supply chain challenges, including expedites, continue to provide cost headwinds, as was also the case with FX this quarter. While we are seeing an impact from the rise in commodity and labor costs, we have also been adjusting pricing, which should help to compensate. Overall, as I mentioned in our last call, our P&L continues to benefit from the marginal profitability of each incremental unit with higher fixed cost absorption. As a result of the great progress on margins, volume, and appropriate management of overhead costs, we were able to achieve an operating margin of just under 15%, exceeding the long-term guidance we've laid out previously. On cash, we generated record operating cash flows of $3.1 billion and continued to invest heavily in the build-out of manufacturing, supercharging, and service capacity. We also continue to retire high interest rate debt, including the early settlement of our 2025 senior notes of $1.8 billion during the quarter. As we look forward, we're clearly quite a bit ahead of the pacing required to achieve our target annual growth rate of 50% this year. Q4 production will depend heavily on availability of parts, but we are driving for continued growth. We are also nearing assembly of our first-production cars in Austin and Berlin. It's important to stress that while the first production car is an important milestone, the hardest work lies ahead in the ramp. Please keep in mind that we are pushing the boundaries on new product and manufacturing technologies at these factories, which makes it difficult to predict the exact pace of the ramp. These factories will also weigh on our margins as we work towards volume production. Overall, I am very proud of what the team has accomplished, and I'm excited for our next phase of growth into Q4 and into 2022. The team has done a tremendous job improving our financial health in a short period of time, while also continuing to improve our precision and pace of execution. Thank you.

Thank you very much. And we will now take questions from institutional and retail shareholders that we posted on our website. The first question is, when should we expect the first vehicles to be delivered with 4680 cells?

Yes, thanks, Martin. Earlier next year, from a non-cell perspective, structural battery crash, range, and reliability testing are on track to be complete this quarter. Testing to date has gone well, and the Fremont manufacturing line is on track to support. However, similar to what Lars said before, this is a new architecture, and unknown unknowns may still exist. Our top priority is ensuring quality in what we deliver. And from a cell perspective, we are comfortable with the design maturity and manufacturing readiness, matching the pack timeline I just mentioned.

Thank you very much. The second question from an institutional investor is, do you still expect to start production of the $25,000 model in 2023? What are the biggest hurdles from now until then?

Yeah, we're working on a strategy to increase our production rates as quickly as possible. I think Zach spoke to that well. And we're doing this while trying to add the least amount of incremental complexities to the business. We don't want to add any new vehicles to our lineup when we're generally in a cell-constrained world. While there is still more runway to grow these existing products, we are focused on Model Y expansion in Austin and Berlin, ramping S & X further in Fremont to restore to past levels, while also growing 3 and Y production in Fremont and Shanghai. As we've mentioned before, after Model Y in Austin, our next product launch will be Cybertruck. And that timing, of course, depends on increasing cell capacity, both from our suppliers and through our in-house cells, as well as many other headwinds we face in supply chain, and completing our currently full plate of products on the table.

Thank you very much. The third question is, with FSD Beta training dataset set to explode exponentially as software is released to a wider and wider audience, are there any early takeaways with regards to how quickly versions can iterate and be pushed out from biweekly to weekly or even daily?

At this point, it's not so much about how much data we can collect, but how quickly we can process the data we've collected. This is where Dojo comes in as we mentioned on AI Day with a substantially faster training computer. In Dojo, we will be able to iterate more often than we do now. If, for instance, training takes one day instead of one week, it makes a huge difference in our ability to push out more updates. But realistically, there's a whole lot more that comes into play when iterating software updates. The whole infrastructure from top to bottom, including testing and validation, needs to be set up for faster iteration. So daily updates are not really realistic for now.

Thank you very much. The next question is, can you provide an update on future model development and how much diversity in your fleet will be necessary to achieve 20 million in annual volumes? The best-selling cars in the world today only sell slightly over 1 million units, so is it possible to achieve 20 million units with just S, X, 3, Y, truck and the $25,000 car?

Thanks, Martin. As we've mentioned before, we've seen record growth of both Model 3 and Model Y segments, where Model 3 is currently the bestselling luxury sedan worldwide. As we mentioned at our shareholder meeting, Model Y is poised to be the bestselling vehicle in the world. Tesla continues to break molds in these vehicle segments and we hope to do so with each new product. As we've said publicly, we'll eventually expand the vehicle lineup to get to larger volumes, and we believe that we will need to be in all major segments across small and mid-size, large sedans, SUVs, and trucks to do so. Along with, of course, the massive space of robotaxi.

Thank you very much. The next question, a question from a retail shareholder is, what is Tesla's goal for vehicle production capacity for the four current factories, Fremont, Shanghai, Austin and Berlin by 2024?

Yeah. Thanks, Martin. Our goal as a company here is to grow on an average pace of 50% per year. And so, you can extrapolate that out. There may be some periods of time in which we're well ahead of that. There could be some periods of time, despite best efforts, where we're slightly lower than that. But that remains the long-term goal of the company. In Fremont, we're continuing to push the boundaries of what's possible there. Over the last 12 months, we've done about 430,000 cars of production. And based upon everything that we know in the factory where the bottlenecks are, what the potential is, we're targeting to increase that another 50%. I think that will be a difficult goal, but that's the goal that the internal team has. And they're going to continue to push on that. As we look towards Shanghai, we're continuing to push the boundaries there, and we continue to ramp production there as well. So most recently, the ramp up to Model Y, which was our biggest contributor of volume in Q3. We'll continue to ramp that factory. And in our plans, with time are to keep growing the capacity in that factory. Austin and Berlin are interesting factories because our first iterations of capacity there are on Model Y. But we've intentionally set these factories and locations in which they have a quite significant amount of land and ability to expand. And so, we'll take Model Y at these factories. We're trying to get to 5,000 cars a week as soon as we can. And then we'll continue to push beyond that, potentially even getting to 10,000 cars per week at those factories. And then we will add Cybertruck here in Austin and continue to grow from there. So, our goal is to get to millions of cars per year over the next couple of years. And then ultimately in the long term, be able to achieve 20 million cars per year. We're going to grow as quickly as is feasibly possible with an eye towards a 50% annual growth rate.

Great. Thank you very much. The next question is, what is your view on the tightening regulatory environment for FSD in the investigation and broad data request by NHTSA? Some of the recent nominees to NHTSA have been publicly critical of FSD, including engaging with short-sellers online. How will you manage these environments?

Thanks, Martin. As we have been for years, we always engage with NHTSA and other worldwide regulatory bodies to share our knowledge and to work with them on our approaches on both active and passive safety. There are ongoing regulatory inquiries taking place all the time, and especially on the subjects. FSD, they're at the cutting edge of technology development. During these investigations, my team and myself are always cooperative as much as possible. We expect to embrace the scrutiny of these products and know that the truth about their performance and the innovations our products have will ultimately be all that matters. In the end then and as I've said on previous calls, we take safety as a top priority in all our designs. This is because our primary motivation is coming from a team of incredible engineers designing software and hardware that saves lives and prevents injuries. And doing so, we'll continue to be transparent to the public on how our technology is both developing from an autopilot safety data, the latest of which we just shared in the shareholder update. And you can also see in review a wide variety of customer-post FSD videos on social media.

Just to add to that, I mean, as Lars said, safety is extremely important for Tesla. It's the right thing to do. And if you look at various independent testing and regulatory testing of our products, you can see the work of incredibly talented engineers in the results of those tests. And our goal in developing safety-oriented software around the car is to continue to go beyond what the hardware is able to provide. If you can prevent a crash from happening, that's the safest way to manage this. And I think at a macro level here, what we're seeing, and this is entirely understandable and expected is that the automotive industry is going through a transition from the traditional car as we know it to more computer software-oriented sensor suites around them that can manage things beyond just what the driver manages. And regulatory bodies are understandably interested in understanding how to regulate in this environment. This has no exception to that. So as Lars mentioned here, I think this is a great thing. We're excited to partner. And we'll work collaboratively with all regulatory bodies who want to go on that journey to the transition to a software-oriented vehicle.

Thank you very much. The next question is, service remains an issue with appointments available weeks or even months out. Likewise, supercharger wait has become untenable at some locations. What concrete steps is Tesla taking to improve the customer experience in these two key areas?

I will take the service part of this question. Andrew, you can take the supercharging part. We have seen an increase in service lead times throughout the summer, and there's a couple of things that have contributed to that based upon the information that we had. The first is that I think this is not unique to us: the return to some sense of normalcy in a post-pandemic world has happened, I think, more quickly than most people expected. What we're seeing here is that the number of miles that people are driving has increased. There may have been some demand for service during 2020 or in the early parts of 2021 that customers put off. And so, there's a bit of a catch-up that's occurring that has increased demand for service. At the same time, in the macro-environment here, logistics, moving parts, sourcing parts, has become increasingly more difficult, which is a well-known issue in the world right now, as well as challenges in the labor market. And so, there's this simultaneous increase in demand for service, where the ability to supply that service has been impacted for the reasons I mentioned. And so, we saw an uptick primarily in Europe and North America in service wait times over the course of the summer. And we've been working extremely hard since then to address this. And we've seen our wait times come down. So, this is not the case in every location, but if you think about it from regional average perspectives, we are seeing improvements there. We remain super-focused on adding locations. And so, over the last year, we've grown our physical footprint of service centers by 35%. We've grown our footprint of mobile repair by over 40%. We're also adding staffing as quickly as we can in the areas that are most impacted by the imbalance of supply and demand for service. But I think the most important part about all of this is — and we've said this on calls before — that the best service is no service. And so, we have been incredibly focused as a company, both on the initial quality of our vehicles and reliability of our vehicles. And we've seen pretty substantial improvements in both of those metrics over the long term and over the last couple of quarters. So, it is something that remains on our minds. We monitor this very closely. But hopefully, that's a helpful explanation into the context and what we're doing.

And on the Supercharger side, the Supercharging team monitors congestion and plans expansion to ensure customer experience meets minimal wait times alongside the growth in our vehicle fleet. While we certainly have work to do in expanding capacity in some congested areas, average congestion on the network has decreased over the past 18 months. Nonetheless, we're not standing still. We are executing accelerating expansion plans globally. The network has doubled in the last 18 months, and we are planning to triple it over the next two years. And even so, on an individual site basis, to combat existing congestion more quickly where it is isolated and programmatic, we expedite local relief sites, deploy mobile Superchargers, and we try to introduce pricing strategies that encourage more off-peak usage to avoid waiting.

Thank you very much.

I want to add one thing on supercharging. If you haven't experienced our latest iteration of battery packs that can handle fast charge rates in combination with our 250-kilowatt charging stations, it's pretty incredible. And this is a really important component to supercharging capacity. The faster you can charge, the more charge sessions that you can have on an individual post. That improves other customer experiences as you're going on a long-term journey because your supercharging times are lower.

Yeah. And we've also maintained an ongoing roadmap on software improvements, dynamic routing to avoid busy superchargers; that's actually really helpful. We take the real-time busyness of the stations into account when choosing where to navigate people on their road trip. And beyond that, we're also continuing to improve the Trip Planner itself and how it estimates how much energy people use so it's not too conservative and ask them to charge more than they need to, which is another thing that can delay a total trip.

Great. Thank you very much. The next question is, is Tesla considering any other ideas other than FSD with the real-world AI that can bring additional software revenue to Tesla? If not, can Tesla consider building interesting games around FSD Beta?

Sure. At AI Day we did talk about a potential future where Dojo could be used as a neural net training platform for other companies. It's not a focus of ours today as we are fully subscribed on Dojo with our internal uses. We do expect to continue to improve the in-car experience in the context of FSD.

Great. Thank you. And maybe last question from shareholders is, how has the FSD take rate changed since the introduction of monthly subscription? Are there any plans to increase the FSD pricing as wider release becomes imminent?

I'll take the second part of the question first: we won't be providing any forward-looking commentary on our pricing strategy or what may happen here over the near term. With respect to the first part of the question, it has been an interesting thing for us to study within the company. What I'll say, just as a general statement before I make a couple of specific comments, is that the things that we learned on FSD subscription today are not necessarily all that relevant. This is really more of a platform for when FSD Beta goes into wide release, and the features and functionality become more accessible to more customers. The second thing that I'll note is that if you look at the monthly pricing of FSD subscription, and then you compare that to the cost of rolling the FSD option into your lease or your loan, on a monthly basis the economical way for a customer to enjoy the features of full self-driving is through purchasing it upfront and enrolling it into their financing. And as a result of that, what we've seen in the data is that we are unable to detect a meaningful change in the upfront take rate of FSD when people purchase cars. We have seen quite a bit of activity of folks trying us to experience what the software has to offer and subscribing to it and enjoying it that route. But again, as I said at the beginning, I think what we've seen so far on FSD subscription is not terribly relevant. We'll see how that plays out in the future as we continue to release more features.

Thank you very much. And now let's go to analyst questions. The next question comes from Pierre Ferragu of New Street Research. Pierre, feel free to unmute your mic and go ahead. Pierre, I think your mic is muted. Okay. While Pierre works on that, let's go to the next analyst. The next question comes from Joseph Spak of RBC. Joe, we cannot hear you. Can you click unmute? Okay. While the team is working on that, let's just go back to the shareholders' questions. The next upcoming question is: can Tesla allow FSD to be transferred to another vehicle for a fee, something less than $10,000? Early adopters are paying the price if they want to upgrade their vehicle. You lose value on the trade-in, and now you have to buy in at a higher cost. On that, what we've seen in the data is that we are unable to detect a meaningful change in the upfront take rate of FSD when people purchase cars. We have seen quite a bit of activity of folks trying to experience what the software has to offer and subscribing to it and enjoying it that route. But again, as I said at the beginning, I think what we've seen so far on FSD subscription is not terribly relevant. We'll see how that plays out in the future as we continue to release more features.

I don't think that this is widely known but we actually already do the sentiment of what this question is asking. If you trade in your Tesla to Tesla, there is a difference in price that we pay for a trade-in that has FSD compared to one that doesn't. And so, there's that premium that we pay to repurchase the FSD. That money can then be applied towards the purchase of the new car. So, I just — we hear this feedback quite a bit, we see it on social media, we see it in the forums, etc. And so, this already does exist. We don't call it out explicitly in the trade-in, but potentially we have increased the price of your trade-in as a result. And hopefully, this clears this up because we do actually do that.

Okay. The next question is, Elon said that we would get an update on Cybertruck in November a year ago, but it hasn't happened and we know there are a lot of updates. Will you show off the new and improved Cybertruck?

Yeah. Thanks, Martin. We get a lot of questions on Cybertruck. We've been busy detailing the Cybertruck to achieve the prototype version we shared with customers a while back. As you may have seen recently on social media, we've built a number of alphas and are currently testing those to further mature the design. And while those point out a few key additions like rear steer, there are also a number of smaller or less visible improvements, though the product is largely true to the initial vision. We'll continue to work through the product in the beta stages that we're in now and look to launch that by next year.

Okay. Thank you very much. We just promoted Pierre to a presenter here.

Can you hear me, guys?

Yes, we can hear you.

Amazing. I'm very impressed that you managed to figure that out. Sorry, I had a connection issue. Anyway. Let me ask my question. I actually, I'm very intrigued by what you guys are doing on the insurance front, and you have now in the market in Texas an insurance product for which the premium varies as a function of the safety score of the driver. I'd love to hear about that. You must have some initial data points so what's the market's reaction, what's the update? And from there, can you tell us about how you think you're going to distribute that? Is that going to go through your installed base very easily or is it going to be like a heavy marketing push? And then will you tell us about your expansion plans. What are the next geographies, what's the timing, how fast is that business line likely to grow in the next few years?

Thank you, Pierre. I'm extremely passionate about our insurance product. We have a terrific team here at Tesla who have been spending a lot of time developing this, and probably listening to the call. We're pretty excited so far. At the highest level, we entered the insurance markets unintentionally, I would say. Our customers were coming to us complaining that the price of traditional insurance was too high. And it was reducing the affordability of a Tesla. Part of our journey here at Tesla is we want as many people as possible to be able to afford our products. That's extremely important to achieving the mission of the company. If you look at the price of insurance as a percentage of what somebody's monthly payment is, it's quite high. We spend extreme amount of effort in manufacturing to take $5 out on cost somewhere, or $10 out somewhere else. If we can get $5, $10, $20, $30 out on a monthly payment, you can calculate what that means in terms of reduction of the price of the car if you finance it, and the leverage of improving insurance cost is huge in terms of affordability. And so that's kind of the context by which we stepped into this. As we started to do more research, essentially the tools by which the insurance industry has traditionally calculated risk are optimized based upon the existing data. But the existing data is limited. So they tend to focus on things like marital status or age, or other attributes. Accident history is a good one, etc. What essentially happens here is customers who are low-risk and don't actually file many claims end up overpaying on their insurance relative to their cost. That overpayment then subsidizes riskier customers. As we looked at this and we looked at the data, and because our cars are connected, because they are essentially computers on wheels, there's enormous amounts of data that we have available to us to be able to assess the attributes of a driver who's operating that car, and whether those attributes correlate with safety. Because we do get a signal when a car has been in an accident. So, we've been spending our time looking at hundreds of different variables and also looking at billions of miles of driving history. We've been able to fit a model that is able to predict with decent accuracy the probability of collision over a period of time. The model is not perfect. The model is a function of the data that we have available. That dataset continues to grow. We continue to experiment with new variables, but we do have a model that works pretty well so far. From that model being able to predict frequency of collision, we can then align that against the price curve. And we can have individualized pricing integrated into the car, integrated into the app, integrated into that customer's experience. With the feedback loop back to the customer on how they are driving after every drive, the attributes that they were successful on or unsuccessful on, and the tips of things that they can do to improve their safety. That's what we've developed. We then included the safety score as part of the FSD Beta enrollment program where we have almost 150,000 cars currently using the safety score. And I believe the latest data is over 100 million miles of driving. We've been able to go back and analyze that data and we've learned two things coming from that. The first is that the probability of collision for a customer using safety score versus not is 30% lower. That's a pretty big difference. It means that the product is working and customers are responding to it. The second thing that we've looked at is what is the probability of collision based upon actual data as a function of a driver's safety score. And that is aligning with our models. Most notably, if you're in the top tier of safety compared to lower tiers, there's a multiplicative difference in probability of collision based upon actual data. So, this is a very new and very exciting frontier for us. I know that was long-winded but we spent a lot of time on this and we put a lot of thought into it. Specifically, with respect to the rollout, the insurance industry in the U.S. is intensely regulated, and it's regulated on a state-by-state level. That means that we require regulatory approvals from each individual department of insurance at each state. Texas is the first state that we launched in. I do want to thank the Texas insurance regulators here, you've been great to work with. We have a roadmap of additional states. We will launch the product in those states as we receive regulatory approvals. Our goal is to be in every major market in which we have cars in. We did a soft launch in Texas a week or so ago. And what we're seeing in initial take-rate data is that, if you compare that to what we're seeing in California, we're off to a good start here. So, we're very excited about it. We're excited about individual risk-based pricing. We're excited about the ability for folks to become safer and as a result, save money. And it feeds into our priority as a company of building the safest products in the world.

Yeah. If I can add to that, it's really exciting for the engineering team and for the finance team to take on safety into their world to explore this basis. Thanks for that, Zach.

Thank you. Pierre, do you have a follow-up question?

No. I think I'm fine, Zach. Thanks for taking the time to answer. It's fascinating and very interesting. Yes, I look forward to that.

Thank you very much. The next question comes from Joseph Spak from RBC.

Thanks. Can you hear me?

Yeah, we can hear you.

Perfect. Zach, as you noted, you hit low-teens operating margins. That was your medium-term target. You're there now, despite the number of challenges and not full utilization to plan. How are you thinking about that target now? Does it allow you to go leverage it, either drive price down further to unlock more demand and invest in other initiatives or does that target need to change? And in the longer term, do you have an aspirational gross margin target as the mix of software and hardware changes?

Yeah. We have achieved — we've actually exceeded our long-term guidance on our operating margin target. So, we're very pleased to see that. And as we look out over the next quarter and the next year, there's a number of puts and takes financially for the company. The launch of Austin and Berlin will have ramp inefficiencies there for some period of time until we get those factories up and running. And so that's likely to put some downward pressure on our margins as those factories ramp. Our goals are to ramp those as quickly as possible. But as Andrew mentioned earlier, there are a number of unknown unknowns that we'll need to work through. We are also in this uncertain environment with respect to cost structure. We are seeing costs increase on the commodity side. We're getting feedback from our suppliers as we're seeing ourselves the impact of labor shortages. And then logistics and expedite costs just continue to be a part of our story here and it's uncertain how that will unfold. It's our hope that these things stabilize. Exactly when that happens is difficult to predict. And we have been adjusting pricing in line with those changes in cost. And so, we'll see how that unfolds over the course of the next year. So, it's difficult on gross margin to say where that will go for those reasons. With respect to operating margin, we've been very focused as a company on managing our overhead expenses and operating expenses. Operating expenses as a percentage of revenue has been declining and I expect that trend to continue. I think the net of all of this is hopefully that we continue to make progress on operating margin over the next four or five quarters. As we think forward, the business up until this point has largely been a hardware automotive business with a little bit of software on top of that. As full self-driving matures, as take rates increase, if we are to raise pricing on that, there's considerable upside both on gross margins and operating margin as that comes to light, as the business starts to become more of a mix of a hardware-based company and a software-based company. So, we feel optimistic about the journey — very optimistic about the journey as we look over into the long term, just a little bit difficult over the next four to five quarters. And we'll continue to update on earnings calls as we learn more information. There's just a lot of uncertainty in the world right now.

Thank you, Joe. Do you have a follow-up?

Yes, the question is just — you mentioned LFP acts globally for standard range models. My understanding is that all comes from China. Is that the continued go-forward plan or do you want to have LFP capabilities in other factories around the world?

Yeah. Certainly, our goal is to localize all key parts of the vehicles on the continent, at least the continent, if not closer to where the vehicles are produced. That is our goal and we’re working internally and with our suppliers to accomplish that goal. And not just at the end assembly level, but as far upstream as possible.

Thank you very much. The next question comes from Colin Langham from Wells Fargo.

Sorry. Do you hear me now?

Yes, we can hear you now. Perfect.

Just actually kind of following up on the question before. You mentioned commodities rising, and when I look at a lot of the key raw materials in the battery, cobalt, nickel, lithium, all up 40%. And I know you guys have done a good job of getting long-term contracts to mitigate that impact. Have you seen so far any impact from that spike? And if not, any sense of when that raw material headwind might actually show up or has shown up?

Yeah. We have seen an impact. Our primary exposure right now is around nickel and aluminum, nickel on the cell, aluminum on non-cell. We have a mixture of contracts with various suppliers. In some materials, we contract directly and we have full exposure to price fluctuations. We do have a number of long-term commitments and long-term contracts in place. We also have contracts where there's some amount of cost-sharing based upon the movement of indexes. And so, as these have been moving, some of those costs have been flowing through to us. It's not a substantial amount of cost, but it's not small. As we look towards the next year, I certainly hope it doesn't play out this way, but it's possible that we continue to see more of a cost headwind as a result of these movements. It's difficult to say precisely, but the volatility in the increases is substantial. There are certain suppliers that maybe up to a certain point have been absorbing some of the increase. As contracts expire there, we have to renew and extend them. We'll have to return to negotiations. What we have to do as a company and what we are intensely focused on is we need to be continuing to drive down the cost of our products, which we have been doing. And we have to overcome cost increases that are outside of our control. So whether that's re-sourcing components or redesigning components or finding ways to be more efficient in manufacturing, we have no choice but to continue on that path and be even more aggressive in light of the macroeconomics here.

And diversification. Right. It doesn't need to be nickel or cobalt; there are always other options.

Thank you. Colin, do you have a follow-up question? Okay. Let's go to the next one. The next question comes from Colin Rusch from Oppenheimer.

Hey, guys, can you hear me okay?

Yeah, we can hear you.

Perfect. Could you talk a little bit about your strategy around anode materials and your ability to leverage that into reduction on the cathode side and performance from the anode side?

Sure. I don't know that I am going to get into too many specifics. But one thing I would say is unlike the commodities discussion we just had, the anode materials are not really in the same situation just in terms of what they are; these components are generally stable commodities. There isn't exactly a tit-for-tat or get a better anode, use less cathode. There is a fundamental ratio that you need to maintain for the cell to function. So, zooming out, the primary focus on the anode side that we have is just ensuring that we are able to continue to reduce the cost of the anode without impeding on the long-term cyclability of the product. It can also help with energy density. As you improve the energy density of the anode, you improve the energy density of the cell. Not directly one-to-one, because you have to pack more cathode in as the anode gets better. That's a focus as well but the trade basis is just cycling versus day-one cost.

That's super helpful. And then, just around the vehicle pricing strategy, obviously, there's a lot of flexibility there for some customers and not. Can you just talk a little bit about the process around vehicle pricing and how quickly you expect to change that and adjust as you see some of these commodity prices flow through the cost structure, and you look at demand dynamics for vehicles?

Yeah. Pricing has been a really difficult thing for us over the last couple of quarters. Part of the challenge is — well, the great thing that we're seeing in the space right now is there appears to be quite a profound awakening of desirability for electric vehicles. And to be totally frank, it's caught us a little bit off guard. That awakening and change in consumer sentiment, I'm sure there's lots of reasons that go into it, but folks want to buy an electric car and folks want to buy a Tesla right now. It's very exciting for us. At the same time, we have installed capacity to build more cars but we're constrained by a number of dynamics as we've talked about in great detail. And we are putting in an extreme effort to build as many cars as we possibly can. It's hard to overstate how extreme the efforts are. We're trying as hard as we can to maximize that capacity and to be able to meet the demand that we're receiving. But the net of all of this is that we're not able to increase production capacity fast enough. So, at the same time, we are seeing macroeconomic cost impacts on our structure as we've discussed previously on the call. So, we're trying to think through if somebody orders a car now, it could be delivered in some cases, and depending upon the car and which factory, could be a couple of months, could be a couple of quarters. The timing in which we build that car will be just before that car gets delivered. And what will the world look like at that point? We're trying to think through how the cost structure is evolving. How does pricing need to change with that? What are the supply dynamics in the space? The other thing that I will just note on pricing is that companies change pricing all the time. The difference is that when Tesla changes its pricing, it's extremely transparent, where that's not always the case otherwise. And sometimes our pricing will increase, sometimes our pricing will reduce. Sometimes to the public, our pricing changes may not seem to make logical sense but there is a strategy that we work behind the scenes as we're balancing supply and demand, as we're also trying to balance various shortages on parts, as we're trying to manage wait times. All of that goes into the optimization here.

Okay. Thank you very much. The next question is from Brian Johnson from Barclays.

By the way, great to hear there's a team at Tesla, and not just one person who shows. I wanted to drill down a bit more on FSD. In December of 2020 in an interview in Germany, your leader said that he expected level 5 autonomy within a year, so that would be now. Yet, we look at the progress in FSD and some of the issues you see on YouTube, and it looks very much like a level 2-plus system that requires vigilance. It's a driver-assist system. So, I guess there's three questions. One, what is the timetable to get to level 4 at least capability; we can deal with the regulatory stuff later. Two, what is the criteria for you to release deferred revenue around FSD when it's having a level 2 system that needs monitoring enough to release that deferred revenue? And then three, maybe you could talk a little bit more about how you plan to work with NHTSA who appear to be asking some questions. They have three requests out to you regarding information around the capabilities and data.

Sure. We'll take them in order. It's difficult to be specific on the timelines. The teams are working extremely hard, iterating on every version. We're being extremely transparent through the release of this to public customers who are posting information online. When you are using full self-driving and you're going through the iterations, you can feel the progress. And for those who don't have it in their cars, social media is excellent at getting a sense for how that's progressing. The team is moving quickly, with every iteration, with every update. They're working very hard on that. On your second question about the criteria to release deferred revenue, the way that this works is we have made certain commitments as to what this product can offer at the time that a customer has purchased that. What we have to assess is, have we met those commitments, and is the software widely available to folks that we've made those commitments to within a certain geography. Given that FSD is still currently in the beta phase, it's invitation only and it's limited, we have not deemed that to be appropriate for recognition of deferred revenue. We'll continue to evolve this and monitor it within the finance team to see when we get to the milestones in which we're comfortable recognizing revenue. On the NHTSA question, Lars, do you want to take that?

Sure. As I said earlier, we always cooperate fully with NHTSA and other regulatory bodies in any investigations they may have, particularly related to ADAS systems. When they came out with the standing general order in July, we were quick to respond to that and one of the first and only companies capable of actually meeting the needs of that report. We continue to send that information to them as required, weekly and as it occurs. With the additional investigations, we meet them with great sincerity and we'll work through them one by one to make sure that all the facts come out and NHTSA is well informed about our strategies for both active safety and passive safety. As you may know, we released updates to our airbag and restraint system last week to Model Y using our fleet data. We worked closely with NHTSA on that and they were fully in the loop before we did it. I think these kinds of things will continue to happen in the new regulatory space as Zach discussed, as we move towards a software-based vehicle and we're happy to be part of that journey.

Thank you very much. And the next question comes from Trip Chowdhry.

Thank you. Very good quarter. I had two quick questions. First is regarding the two upcoming factories in Berlin and Austin. How are the two factories different from each other, maybe in the layout, design, assembly lines? And the second question is related to Cybertruck. Who is supplying the exoskeleton steel? Is the supply for that material sufficient for immediate ramp up, say in 2023-24 time for Cybertruck? That's all for me.

Yes. So obviously, as we've noted in the past, we developed our own stainless-steel grade for the exterior of Cybertruck to meet both the durability and corrosion requirements required for an automotive world. With this raw material and others, as Andrew mentioned, we continue to look at multiple sources. We have made some early sourcing decisions in that but I think we'll keep some of those internal and we've already begun the first sourcing in. Rolling stainless isn't so different from rolling any other materials. It's about how hard the rollers are to get to that hardness level. Just like every manufacturing process we put in for every new vehicle, we'll work with our suppliers and vendors to make sure those timelines and supply meet the needs and demand of our customers.

And on the differences between Austin and Berlin, there are some. They're largely associated with the different building architectural choices that occurred in collaboration with local codes and other governing requirements that drive the differences in the architecture between the locations. In general, though, we're trying to progress the manufacturing system as a system and make a logical path to find improvements from factory to factory. In some cases, there was an improvement identified between one and the other. So there might be a slightly newer iteration of one part of the factory in one place than the other, but it's all part of a path forward in the factory that builds the machine that builds the machine.

Thank you very much. And the last question comes from Jed Dorsheimer from Canaccord.

Can you hear me?

Yes, we can hear you.

Hey, thanks for taking my question. So, Brandenburg, I'm just wondering, Zach, if you could estimate the carry costs from a margin perspective. Or I guess in two parts. One, do you still expect production coming on in '21? So, a couple of months left in December. And how do you see that margin impact as a function of the carry costs? I do have a follow-up.

Sure. So, it remains our target in both Austin and Berlin to be able to build our first production cars before the end of the year. We've talked about this a bit. The unknown unknowns — new factories, new vehicle designs, new technologies, new locations, new teams — mean there is quite an execution journey ahead of us, but that remains our target. All of our plans are oriented around that. We should not expect for us to deliver cars by the end of 2021 from these factories to customers, even if we do produce some. Homologation and regulatory reasons apply, and we want to make sure that we build up some number of cars that we're confident in the quality and the customer experience around them. The second thing I'll say and I mentioned this in my opening remarks is because of the newness here, it's extremely difficult for us to be precise in what the ramp will look like. It's possible the stars align and things move quickly. It's possible that we're spending the bulk of next year working on ramping these factories. It's just very hard to say, and we'll continue to update you on these calls and through other forums. As to how that then impacts our margins, that is also difficult because that is a function of the ramp, which is uncertain. The benefit here, which is different in the ramp of these factories compared to other factories, is if you think about the percentage of our total cost structure in any given quarter that is associated with new ramps, we have the Fremont factory that's running — generating stable and growing margins there. The same is also true in Shanghai. I expect we'll see some amount of headwind on margin from these ramps. It's entirely dependent on how quickly we're able to ramp, and what uncertainties come up during the process.

Sure, so margins per car, I would suspect though, if your carry cost is full rate now, then as you start producing vehicles, it's going to be a margin lifter from where you're at right now, no?

We are carrying some amount of costs associated with the factories today. So the incremental costs associated with turning the factory is not 100% of the factory, if that's what you're getting at. We also actually saw a very similar dynamic to this when we were relaunching Model S earlier in the year. When a product starts launching and then cost of goods sold starts to activate, depreciation starts to activate, there's a bit of a movement in the P&L as to where that cost resides. So yes, to some extent, Brandenburg and Austin costs are already flowing through our P&L, but we still need to continue staffing and ramping, incurring all the operating costs associated with the factory that we're not spending right now.

Fantastic. Thank you very much, everyone, for all your questions. And we'll see you again in three months. Thank you very much and goodbye.